Target State Estimation Method in Dynamic Environment

Abstract

The present invention relates to a target state prediction method in a dynamic environment. For different colors in the dynamic environment, a color space with a best identification effect is extracted, then through continuously updating a next frame center point and a next frame image threshold range, an effect of adapting to target detection is achieved, threshold range estimation can be effectively shortened, a target is identified in a maximum way, through the combination of HSV and YCbCr color space, an update center point is corrected and restored to improve the accuracy of an algorithm, then the prediction of a target is tracked according to a Kalman filter and a Markoff model, the interference of discrete color features and similar objects in a complex environment is overcome, a detector search range is reduced, the tracking precision in the partial occlusion of the target is improved, and the tracking accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of image processing and pattern recognition, in particular to a target state estimation method in a dynamic environment. Background technique [0002] How to extract the object of our interest from an image and always calibrate the target position in the image sequence is the main topic of target tracking, and it is also an enduring topic in image processing. The earlier research in this field and the more widely used image segmentation technology is suitable for image target extraction. It is a classic problem in the field of image processing and analysis, and it is also one of the difficulties in this field. Image segmentation is actually a division problem. According to specific division criteria, the pixels in the image are filtered and divided. Usually, the result of division is to distinguish the background from the extract, or to highlight the extract, or to eliminate noise. Through division, the Ima...

AI Technical Summary

Problems solved by technology

[0004] In target recognition methods, classic target recognition methods include Otsu (inter-class variance method), maximum entropy method, basic global threshold method, iterative threshold segmentation method and various adaptive threshold methods, all of which are based on grayscale images. Single-channel segmentation method, this method can achieve a good segmentation effect in the case of large differences between the target and the background and low noise, but the tracking effect of the target with complex background and rich texture features will be relatively poor, and it is easy to be affected by the background. The interference of objects with similar targets will eventually lead to the failure of target recognition and tracking

Images